Torbjorn Zetterlund

Tue 16 2024

Business vs. GPTs: Navigating the Challenges of Personalized AI

by bernt & torsten

The recent proclamation by OpenAI has caught the attention of $20 ChatGPT Plus subscribers, who now can master their own personal Generation Pre-trained Transformers (GPTs). With this newfound power, individuals with interests ranging from yoga to SEO have started tailoring these GPTs to their liking.

In an exciting development, OpenAI is gearing up to launch a GPT Store, demonstrating the expansive potential of this technology. The latest version is set to offer groundbreaking features, including the ability to upload personal knowledge, carry out dynamic code execution, and generate images.

Adopting this innovative technology has its pitfalls for businesses. One of the primary challenges revolves around the fact that any data uploaded to the GPTs could become part of ChatGPT’s training data, thereby posing a risk to the confidentiality of the information. There are also fears about data privacy, given the possibility that competitors could download raw files.

Moreover, the time-intensive process of crafting custom instructions is constantly threatened by leaks to competitors or other creators, which could harm the uniqueness of a GPT. Complicating matters further, there needs to be a more convenient way to embed GPTs into websites, entailing that businesses would have to attract users to an external site for a subscription to ChatGPT Plus.

There is a need for built-in measures to prevent the generation of wrong information, thereby raising possible legal and brand-related issues for businesses looking for reliable responses. Also, the lack of access to chat logs hampers businesses’ efforts to understand user behaviour or improve GPTs in the long run, thus transforming it into a sort of high-risk ‘black box.’

Adding to these are practical limitations such as the binding limit on knowledge base size and word length and the inhibiting inability to incorporate diverse data formats. These drastically reduce the utility of GPTs for business applications. Further, GPTs have shown difficulty in effectively digesting website data, limiting their usefulness as a comprehensive information source.

The procedural bottleneck also comes from restricting accessibility to paid subscribers, which bars businesses from using GPTs for customer service, lead generation, and other vital functions. As we can see, while the OpenAI GPT Builder presents immense potential, significant hurdles must be surmounted to become a viable tool for businesses without resorting to complex developmental projects and infrastructural management.